作者你好，你能给个训练过程中几个loss的收敛情况吗？

SystemErrorWang / White-box-Cartoonization

Official tensorflow implementation for CVPR2020 paper “Learning to Cartoonize Using White-box Cartoon Representations”

3.93k stars 737 forks source link

Open ersanliqiao opened 3 years ago

ersanliqiao commented 3 years ago

我这边训练下来，判别器比较稳定下降，但是g_loss_total和photo_loss一直增加

ersanliqiao commented 3 years ago

SystemErrorWang commented 3 years ago

@ersanliqiao 是正常的，判别器loss基本不动，生成器loss增大是因为卡通化效果越来越明显，和真实图片相似度就会越来越小